Polit Behav (2013) 35:777–806
DOI 10.1007/s11109-012-9217-x
ORIGINAL PAPER
Neighborhood Disorder and Local Participation:
Examining the Political Relevance of ‘‘Broken
Windows’’
Jamila Michener
Published online: 24 December 2012
! Springer Science+Business Media New York 2012
Abstract Anyone who has lived in, driven through or walked by a ‘‘bad’’
neighborhood has a sense of the attributes that render such places unique: graffiti,
litter, public intoxication and much more. According to the well-known theory of
‘‘broken windows,’’ these readily observable corporeal characteristics signal
neighborhood disorder and lead to increased criminal behavior. This article investigates the implications of disorder for political behavior, taking particular care to
distinguish between the objective tangible conditions of disorder and residents’
subjective interpretations of those conditions. Utilizing exceptionally rich data, this
analysis reveals that while certain aspects of objective ‘‘reality’’ are consequential,
perceptions of such reality are a more powerful mechanism through which neighborhood disorder impacts local political engagement. For some political outcomes, a
heightened sense of the problems associated with disorder is linearly associated with
an increase in participation. For others, the pattern is parabolic: those who perceive
so little disorder that they remain content or so much disorder that they become
disaffected are substantially less likely to take action to make their communities
better. Ultimately, holding objective contextual features constant, the lenses through
which residents interpret things like ‘‘broken windows’’ are critical determinants of
grassroots politics. This information, combined with broader understandings of what
shapes perceptions of disorder, lays the foundation for structuring policy in ways
that facilitate grassroots activism—a vital component of the American democratic
process.
J. Michener (&)
University of Michigan, 1415 Washington Heights, M2232, Ann Arbor, MI 48109, USA
e-mail: [email protected]
J. Michener
Cornell University, Ithaca, NY 14850, USA
123
778
Polit Behav (2013) 35:777–806
Keywords Perceptions ! Neighborhood ! Political participation !
Broken windows ! Disorder ! Objective context ! Subjective context !
Community engagement ! Local politics
Anyone who has lived in, driven through, worked near or walked by a ‘‘bad’’
neighborhood has a sense of the tangible attributes that render such places unique:
graffiti, litter, abandoned buildings, public intoxication and much more. These
corporeal characteristics provoke strong, sometimes visceral reactions from
residents and passersby. In fact, according to the popular theory of ‘‘broken
windows,’’ the directly observable material conditions of a neighborhood have an
influence on the ways people respond to it. Proponents of this theory argue that
minor but visible signs of neighborhood disorder1 lead to more serious criminal
violations (Wilson and Kelling 1982; Kelling and Coles 1996). Scholarly and public
reactions to this claim have been numerous and contentious, sparking much
argument over whether and how broken windows theory should inform policies on
crime and punishment (Lott 2000; Levitt and Dubner 2005; Brook 2006; Harcourt
and Ludwig 2006; Skogan 2008; Roberts 2011).
Expanding beyond the criminological roots of this debate, social scientists across a
range of disciplines have begun to examine the broader consequences of disorder,
arguing that it, ‘‘plays a role in undermining the stability of urban neighborhoods,
undercutting natural processes of informal social control, discouraging investment
and stimulating fear of crime’’ (Skogan 2012: 174). In short, accumulating evidence
points to the multifaceted effects of disorder on social, economic and psychological
outcomes (Casciano and Massey 2011; Hill et al. 2005; Christie-Mizell and Erickson
2007). Despite this literature’s growing reach, scholars have remained silent about the
political relevance of disorder. This is especially problematic because the possibility
of transforming blighted neighborhoods is at least partially rooted in local political
engagement (Florin 1989; Putnam 1993; Marschall 2004; Epstein et al. 2006). Without
knowing the ramifications of disorder for local politics, we cannot adequately map the
participatory pathways that might enable urban dwellers to forefend against it.
This article draws on the core intuition of broken windows theory—that salient
physical and social neighborhood conditions impact behavior—in order to examine the
relationship between local political participation and neighborhood disorder. Utilizing
exceptionally rich data that includes both systematically observed objective markers of
disorder and survey-based subjective perceptions of it, this analysis reveals that while
certain aspects of objective ‘‘reality’’ are politically consequential, perceptions of such
1
There is much contention among sociologists and criminologists about how best to define the concept
of disorder (Harcourt 2001; Kubrin 2008). I will not attempt to resolve this issue here. I agree that there
are problems with the idea (many of which are succinctly described by Kubrin 2008) but nonetheless
contend that it captures (however imperfectly) a phenomena that is important for conceptualizing
neighborhood context and is therefore worthy of study by political scientists. With this in mind, I take a
practical approach and follow Skogan (2012: 174) in characterizing disorder as including, ‘‘unsettling or
potentially threatening and perhaps unlawful public behaviors’’ (social) and ‘‘overt signs of negligence or
unchecked decay as well as the visible consequences of malevolent misconduct’’ (physical). This
definition is broadly reflective of the approach that has been taken in the literature to date.
123
Polit Behav (2013) 35:777–806
779
reality are a more powerful mechanism through which neighborhood disorder affects
local political engagement. Perceptions of disorder have robust linear and curvilinear
relationships with participation. In some cases, a heightened sense of the problems
accompanying disorder is linearly associated with an increased likelihood to
participate. In other cases the pattern is parabolic: those who perceive so little disorder
that they remain highly content or so much disorder that they are sharply disaffected,
are substantially less likely to take action to make their communities better places to
live. Ultimately, even holding constant objective neighborhood features, perceptions of
disorder are critical determinants of grassroots politics.
Various implications stem from these findings. As I discuss in more detail later,
the policy levers and organizational strategies necessary for cultivating community
participation can be seen in a different light when perceptions become a more salient
emphasis. Through a lens focused on perceptions, proposed solutions to urban crime
or neighborhood economic disadvantage are not only relevant vis-à-vis their direct
bearing on lived conditions, but also because of their influence on how everyday
people interpret their neighborhoods and (by extension) how they make decisions
about whether to invest their political energies into solving local problems.
In addition to its potential policy applications, this research also has substantial
scholarly value. It brings heightened conceptual clarity to the study of contextual
effects by theoretically and empirically delineating the significance of the
perception-reality distinction and applying it to political outcomes. To date, much
of the work on neighborhood effects has proceeded without fully accounting for
either the material physical qualities of neighborhoods or the varied subjective
interpretations of such qualities. This too often has detached the inquiries of social
scientists from the realities of ordinary people. The present research bridges that
gulf by examining neighborhood disorder on both experiential and perceptual levels.
Objective Context and Subjective Perceptions: Why Make the Distinction?
The importance of physical and social environments to political behavior emerged
as a theme in the study of American politics as early as the 1930s and has flourished
since (Hauser 1974; Huckfeldt 1986; Huckfeldt and Sprague 1993; Cohen and
Dawson 1993; Burbank 1995; Alex-Assensoh 1998; Oliver 2000; Anderson 2009;
Gay 2012). Despite this profusion of first-rate research, political scientists have too
often neglected the duality embedded in the concept of a neighborhood (Burbank
1995; Sampson and Raudenbush 2004; Santiago et al. 2010; Meersman 2005;
Quillian and Pager 2001; Hadley-Ives et al. 2000; Ross 2000). As Lee and Campbell
(1997) point out:
…most neighborhoods… are social constructions with an existence rooted in
residents’ awareness as well as in particular physical settings. Their
ambiguous character is nicely captured by the notion of a ‘quasi-fact,’’ a
phenomenon that lies at the juncture of the subjective and objective realms.
Studies of neighborhood effects have largely overlooked this quality and have
instead attempted to capture contextual influence by including census or other
123
780
Polit Behav (2013) 35:777–806
‘‘objective’’ data about the demographic composition of neighborhoods (e.g. percent
in poverty). While such patterns are imperative, they do not adequately reflect what
people encounter on their blocks everyday nor do they have uniform meanings
across or within neighborhoods. As a result, demographics fall short of
encompassing the assorted factors that account for the relationship between
neighborhoods and political behavior.
Perhaps as a consequence of over reliance on neighborhood demographic measures,
much of the contextual effects research has also lacked thoughtful inclusion of
subjective understandings of neighborhood. Instead, researchers have tended to ignore
or misestimate the role of perceptions (Books and Prysby 1988; Eagles 1995). Models
of contextual effects recurrently include survey-based measures of neighborhood
perceptions that vary widely in their content and theoretical meaning: sometimes they
are intended to replace objective features that have not been explicitly modeled, other
times they are the main channel through which neighborhood effects operate and many
times there is no overt mention of their larger conceptual status.
Lacking satisfactory understandings of either objective neighborhood realities or
subjective perceptions of those realities, many researchers overlook the question of
whether and how the two are related. Those who do pay attention, particularly
scholars in the fields of psychology and health, consistently find low to moderate
agreement between objective measures of neighborhood and subjective perceptions
of those same environments (Kamphuis et al. 2010; Kweon et al. 2006; Hadley-Ives
et al. 2000; Ross 2000; Aneshensel and Sucoff 1996). In short, the existing evidence
indicates that the concrete material conditions of a neighborhood and the subjective
perceptions of such conditions are related yet very distinct phenomenon.
From the perspective of psychologists who study perception, this makes perfect
sense. Theories of cognitive mapping have established that mental representations
of the environment differ markedly from its physically measurable components
(Brunswik 1956; Kaplan and Kaplan1982). Psychologists explain this by emphasizing the function of perception, which is much more than simply seeing what is
‘‘really’’ out there. Instead, it involves the organization and interpretation of sensory
information such that even when a stimulus is held constant, there can be stark
variations in how it is perceived conditional on the other aspects of the perceptual
process (Pomerantz 2003).
Notwithstanding the complexity of this process, a handful of studies across
various disciplines have scrutinized the connection between perception and reality
as it regards the proper measurement of neighborhood effects. The basic consensus
emerging from this literature is that objective context is a consistent determinant of
perceptions, though only one of many and often not the strongest (Kamphuis et al.
2010; Franzini et al. 2008; Christie-Mizell and Erickson 2007; Sampson and
Raudenbush 2004). Further still, while the task of appropriately gauging the effect
of neighborhood context is best achieved by accounting for both objective and
subjective measures, perceptions often have a larger direct effect than do objective
factors (Aneshensel and Sucoff 1996; Christie-Mizell and Erickson 2007; Latkin
and Curry 2003; Weden et al. 2008).
Notably, none of these insights are drawn from research that examines political
behavior. Keeping this in mind, I glean knowledge from this literature and apply it
123
Polit Behav (2013) 35:777–806
781
Conceptual Model
Neighborhood Context
Objective Disorder
(Systematic Social
Observations)
G
A
C
Subjective
Disorder
(Perceptions of
Neighborhood)
F
D
Local Political
Participation
B
Demographic
(Aggregate
Neighborhood Cluster
Data)
Individual
Characteristics
E
Fig. 1 This figure illustrates the conceptual model motivating the hypotheses about the relationship
between neighborhood context and political participation. Note that subjective perceptions are strongly
hypothesized to have a direct effect on participation (line D) and that objective disorder is weakly
hypothesized to have a direct effect on participation (dotted line G)
to the political setting in order to clarify how neighborhood disorder influences local
political participation.
Conceptual Model and Hypotheses
A first step in understanding the associations between objective context, subjective
perceptions and participation is to properly specify the various relationships at work.
To this end, I propose a conceptual model (see Fig. 1). As shown, the model
suggests that disorder (both objective and subjective) along with demographic
features and individual characteristics are key predictors of local participation.
Among these, political scientists have carefully examined the participatory
significance of individual (line F) and demographic (line E) factors (Huckfeldt
1986; Huckfeldt and Sprague 1993; Rosenstone and Hansen 1993; Verba et al.
1995; Cohen and Dawson 1993). Similarly, sociologists have been attentive to the
relative importance of individual (line C), demographic (line B) and objective (line
A) factors in shaping subjective perceptions (Sampson and Raudenbush 20042;
Franzini et al. 2008; Elo et al. 2009). The only relationships in this model that have
2
In fact, Sampson and Raudenbush (2004) use the data leveraged in this paper to establish the
connection between subjective, objective, individual and demographic factors.
123
782
Polit Behav (2013) 35:777–806
not been tested by social scientists are those that establish the influence of subjective
and objective disorder on participation (line D and dotted line G, respectively).
In scrutinizing these relationships, I test two hypotheses. The first, H1, is a
perceptions-oriented hypothesis, which asserts that subjective perceptions of
disorder will have a direct effect on local participation (line D). H1 is rooted in
the intuition that residents’ sense of the severity of disorder in their immediate
surroundings will shape their inclination to invest time and energy into local
political activities. There is an abundance of empirical evidence indicating that
perceptions of neighborhood and local participation are closely related (Kasl and
Harburg 1972; Chavis and Wandersman 1990; Taylor 2001; Small 2004; Manzo and
Perkins 2006). Manzo and Perkins (2006: 336) insist that, ‘‘Our thoughts, feelings,
and beliefs about our local community places…impact our behaviors toward such
places, thus influencing whether and how we might participate.’’ Small (2004: xv)
reinforces the notion that heterogeneous perceptual responses to neighborhoods
likely have consequences asserting that, ‘‘…not everyone sees the same neighborhood through the same eyes [and] how residents see their neighborhood affects how
they react to it and whether they are willing to ‘get involved.’’’
The research cited in favor of H1 has not expressly investigated the participatory
outcomes associated with perceptions of disorder (as opposed to more general
feelings about the neighborhood), thus omitting a uniquely important perceptual
category. Despite this absence, the sociological and psychological literatures gives
us strong reason for believing that perceptions of disorder, like their more general
counterparts, should matter. As such, I expect that H1 will be supported by the
findings that follow.
The second hypothesis that lies at the core of the paper, H2, is a reality-oriented
hypothesis, which posits that objective disorder will have a direct effect on local
participation (dotted line G). Previous studies strongly support the claim for an
indirect effect of objective disorder (via perceptions) by consistently showing that
such disorder shapes subjective perceptions (Sampson and Raudenbush 2004;
Christie-Mizell and Erickson 2007; Franzini et al. 2008; Elo et al. 2009; Kamphuis
et al. 2010). This has been well enough established in the literature that I do not test
it here. If H1 proves true, then we have good reason to suppose that objective
disorder influences participation indirectly by shaping perceptions of disorder. The
more unsettled matter, however, involves the direct consequence of objective
disorder. To date, the research record is void of evidence about the connection
between objective disorder and local participation. Further still, ordinary deductive
reasoning can supply accounts both for and against a direct relationship.
On the one hand, even per the original broken windows theory, a key aspect of
disorder is its visibility, which presumably links disorder to behavior via signaling
(Wilson and Kelling 1982). To the extent that a signal must be perceived in order to
be effectual, perception should be the primary avenue through which objective
disorder indirectly operates.
At the same time, there is at least face validity to the alternative prospect that
objective neighborhood disorder has a direct effect on political participation through
channels separate from perception. Weden et al. (2008: 1258) make an argument
that buttresses this scenario in saying that, ‘‘objective measures may capture
123
Polit Behav (2013) 35:777–806
783
important structural aspects of the environment that the respondent may not or
cannot perceive.’’ They give the example that neighborhood disadvantage may alter
access to local resources and consequently affect other outcomes, even if people do
not directly perceive these associations at work. In a case like this, disorder might
function irrespective of signaling.
Ultimately, inadequate empirical verification dampens the (a priori) strength of
our confidence in H2 (hence the dotted line G). The models that follow will provide
an evidentiary basis for either fortifying or further reducing such confidence going
forward.
Before moving on, it is worth noting two things. First, H1 and H2 are not
mutually exclusive. While the above assertions position H1 as the stronger of the
two postulates, it is entirely possible that both will be confirmed. Second, I
purposely say nothing about the direction (e.g. positive, negative) of the
relationships between subjective perception and political participation. This is
because the research record is mixed in this regard. Much existing scholarship links
negative perceptions of neighborhood to depressed participation. In an in-depth
exploration of community participation in a Boston neighborhood called Villa
Victoria, Mario Small discovered that older cohorts of neighborhood residents who
perceived Villa Victoria as, ‘‘a beautiful place to live’’ were more active participants
in neighborhood activities than their younger counterparts who had less positive
views of the neighborhood. Small identified, ‘‘a shift in the perception of the
neighborhood by residents’’ as a ‘‘mechanism by which participation declined in
Villa Victoria’’ (Small 2004: 87). Other studies also buttress this logic: Ross et al.
(2001) find that perceived neighborhood disorder is associated with feelings of
powerlessness; Manzo and Perkins (2006) find that perceptions of inclivities can
lead to decreased involvement in community organizations; Rohe and Basolo
(1997) find that as the perception of neighborhood improves people are more likely
to attend social and political meetings.
Yet, there are at least two alternatives that also have some backing in the literature.
First, negative perceptions may mobilize residents, leading them to become more
involved in efforts to improve local conditions (Haney 2007). Second, the relationship
between perceptions and participation may be non-linear, going in one direction under
certain conditions and another direction under others (Wandersman et al. 1981). Given
the indeterminacy of the literature, H1 underscores that perceptions are important but
remains neutral with respect to the direction of the effect.
Data
In elucidating the relationship between disorder (objective and subjective) and
participation, I integrate three types of data: (1) individual level survey data
measuring perceptions of disorder and other individual attributes, (2) systematic
social observations (SSO) providing independent assessments of objective disorder,
(3) aggregate data gauging the basic demographic makeup of neighborhoods. All of
the data is drawn from the 1995 Project on Human Development in Chicago
Neighborhoods (PHDCN).
123
784
Polit Behav (2013) 35:777–806
One potential drawback of the PHDCN data is that it is limited to Chicago. Yet,
while the truncated range of the data is a weakness of breadth, it is a strength of
depth. Narrowing the scale of the data collection made it feasible to gather
information that is extraordinarily rich. Moreover, even standing alone, findings
from PHDCN data can tell us about patterns in one of the largest urban centers in
the country, thus verifying whether the variables under scrutiny are relevant, even if
not universally so (the latter claim would be fodder for future research).
Survey Data
The individual level data comes from the PHDCN Community Survey.3 This survey
involved a multistage probability sample of 8,782 Chicago residents. The first step
in the sampling process was to delineate 343 ‘‘neighborhood clusters,’’ (i.e.
aggregated census tracts). The predominant criteria in the formation of the
Neighborhood Clusters (NCs) was that they were ecologically meaningful,
composed of geographically contiguous census tracts, internally homogenous on
key census indicators (i.e. race and class) and informed by local geographic
knowledge.4 These are not neighborhoods per se, but they are geographic units that
can feasibly be conceptualized as communities both from the perspective of the
people living there and in terms of demographic homogeneity. Once the NCs were
established the sampling proceeded in stages: city blocks were sampled within each
NC, dwelling units were sampled within blocks and one adult resident was sampled
within each dwelling unit.
There were three sets of variables culled from the individual level survey data:
(1) political participation variables, (2) perceptions of physical and social
neighborhood disorder, (3) important individual level controls. Summary statistics
are provided in Table 1.
Political Participation Measures
Perhaps as a testament to its sociological lineage, the PHDCN community survey
does not contain the most traditional measures of political participation such as
voting, registration, candidate choice or mainstream political attitudes. Even so,
there are at least two variables that involve political action on the local level, a
phenomenon we would expect to be most proximately impacted by neighborhood
disorder. Specifically, respondents were asked whether they had: (1) spoken to a
politician about a neighborhood problem, (2) attended a meeting of a block or
neighborhood group about a neighborhood problem. While these dependent
variables have a moderately high correlation with one another (.60), they represent
distinct political behaviors. The first outcome is connected to the formal political
system while the second taps into a broader kind of political participation that
happens outside of the purview of the sate. While similar proportions of the sampled
3
Earls et al. (1997).
4
See the following for more details on the sampling methods associated with PHDCN: http://www.icpsr.
umich.edu/PHDCN/sampling.html.
123
Polit Behav (2013) 35:777–806
785
Table 1 Summary Statistics
Variable
Min
Max
Mean
Standard
deviation
Individual level
Spoke to politician
0 (=no)
1 (=yes)
.29
.45
Attended meeting
0 (=no)
1 (=yes)
.27
.44
Age
17
100
42.7
16.7
Sex
0 (=male)
1 (=female)
.59
.49
Latino
0 (=no)
1 (=yes)
.25
.43
African-American
0 (=no)
1 (=yes)
.39
.49
Income
1 (\$5 K)
15 ([150 K)
5.7
3.5
Education (years of
education)
0
17?
12.3
3.9
Mobility (# times
moved)
0
11
.95
1.4
Home owner
0 (No)
1 (Yes)
.45
.50
Perceptions of
disorder
6 (=fewest perceptions of
disorder in neighborhood)
18 (=most perceptions of
disorder in
neighborhood)
10.8
3.9
Physical disorder
0
17.5
5.4
2.2
Social disorder
0
5
.14
.37
Income
1 (\$5 K)
13 ($100–124.9 K)
5.7
1.7
Black
0
1
.39
.40
Tenure length (years
in neighborhood)
2
24
10.3
4.4
Neighborhood Level
population engage in each of these activities (29 percent report having spoken to a
politician and 27 percent report having attended a meeting), not all of the same
people who participated in one activity also participated in the other. Substantial
minorities of the sample picked one form of participation over the other: 33 percent
of the people who attended a meeting, did not speak to a politician while nearly 38
percent of those who spoke to a politician, did not attend a meeting. In the event that
the overlapping but distinct subsets of people engaging in these political activities
mattered for the questions of interest in the paper, I modeled them separately.
Perception Measures
In order to gauge perceptions of physical and social neighborhood disorder, I relied
on six pertinent questions about how much of a problem respondents saw the
following in their neighborhood: (1) litter/trash, (2) graffiti, (3) vacant housing/
storefronts, (4) drinking in public, (5) selling or using drugs, (6) teenagers causing a
disturbance. Respondents could rate each of these phenomenon on a three point
scale where 1 = not a problem 2 = somewhat of a problem and 3 = a big problem.
To make the analysis more tractable, I used these six perceptual items to construct a
123
786
Polit Behav (2013) 35:777–806
composite scale that ranged from 6 to 18, where increasing scores on the scale
represented increased perceptions of disorder (Cronbach a = .86).5
Demographic/Control Measures
Seven measures were included as individual level controls: age, education,
household income, a dummy variable indicating whether a respondent was Latino,
a dummy indicating whether a respondent was Black, a measure of mobility
(number of times the respondent had moved in past five years) and an indicator for
home ownership. While there were many more variables available, I limited the
controls to those most grounded in the literature (Rosenstone and Hansen 1993;
Verba et al. 1995; Bolan 1997; Lake and Huckfeldt 1998; Sampson et al. 2002;
Aneshensel and Sucoff 1996; Christie-Mizell and Erickson 2007; Sampson and
Raudenbush 2004; Rohe and Stegman 1994).
Systematic Social Observation Data
The second source of data was the Systematic Social Observation (SSO) data, also
collected for the PHDCN.6 SSO is a standardized method for directly observing the
physical and social characteristics of neighborhoods, one block at a time. To
assemble this data, trained employees of NORC were given explicit rules for
recording and observing visual physical and social characteristics of various
Chicago neighborhoods. More specifically, a driver, videographer and two trained
observers drove an SUV at less than 5 miles per hour down every street of 196
Chicago Census tracts.7 As they drove, video cameras on each side of the vehicle
recorded the social activities and physical features of each side of the block while
the observers simultaneously took a log of what they saw. The observation took
place on all days of the week from 7 a.m. to 7 p.m. over the course of 6 months. In
total, 23,816 face blocks (e.g. one side of the street) were observed. Trained coders
coded all observer logs along with a random sample of over 15,000 face block
videotapes. As a check on reliability, new coders randomly recoded 10 percent of all
the data and produced codes with 98 percent inter-coder reliability (Sampson and
Raudenbush 1999, 2004).
The data generated from this observational process was vast. It included
information on land use, traffic, the physical condition of streets and buildings, the
presence or absence of various types of neighborhood institutions and much more
5
I considered creating two scales: one measuring perceptions of physical disorder and another gauging
perceptions of social disorder. My primary reason for not doing so was theoretical. I had no strong
theoretical impetus for assuming that individuals’ interpretations of their neighborhoods are filtered via
such a dichotomy. Furthermore, a principal components analysis revealed that the combined items
reflected one latent variable. More specifically, the first component had an Eigen value of 3.6 and
captured over 60 percent of the variation between scale items. Finally, to be sure that this measurement
decision did not impact the subsequent analyses, I re-ran the main models using two separate perceptions
scales and found little difference in the results.
6
Earls et al. (2002).
7
These tracts were chosen based on a stratified probability sample designed to maximize race and class
variation.
123
Polit Behav (2013) 35:777–806
787
(Sampson and Raudenbush 1999). For the purposes of this project, I followed
Sampson and Raudenbush (2004) in constructing two theoretically motivated SSO
scales measuring physical and social disorder. The physical disorder scale consisted
of 7 items gauging the material presence of: (1) beer bottles, (2) litter, (3) syringes,
(4) cigarettes, (5) condoms, (6) abandoned cars and (7) graffiti. The social disorder
scale measured the following five activities: (1) adults loitering, (2) open alcohol
consumption, (3) drugs being sold, (4) intoxicated persons on the block face, (5)
prostitution.8 The scales were coded such that higher scores indicate increased
observations of disorder. Arguably, each scale taps into substantively different
phenomenon as evinced by their low correlation with one another (.10).9
Neighborhood Demographic Data
In order to control for neighborhood level demographic patterns, I aggregated data
from the individual level survey to the Neighborhood Cluster level.10 The
demographic controls include measures of mean income, percent black, and mean
tenure length (e.g. average number of years NC residents lived at their current
address).
Analytical Strategy and Statistical Models
Given the use of observational data and all of the limitations that come with it, the
subsequent analyses include four strategies meant to make the findings as robust as
possible. First, I include a range of basic controls to avoid omitting critical
variables. Second, I address the nested structure of the data via multilevel modeling.
Third, I account for non-linear patterns. Finally, I conduct robustness checks using
different model specifications to confirm that the key findings are not spurious or
idiosyncratic.
I test H1 and H2 via multilevel logistic regression. Hierarchical analyses are
uniquely suited to questions that involve multilevel effects (Bryk and Raudenbush
2002; Steenbergen and Jones 2002). In the equations specified, the individual level
8
For more information about how each of these phenomenon were coded see Sampson and Raudenbush
(1999): ‘‘Systematic Social Observation of Public Spaces: A New Look at Disorder in Urban
Neighborhoods.’’
9
In addition to their low correlation, both scales had relatively high Eigen values (above 1). However, it
is important to note for these measures in particular that the construction of the scales was driven more by
theory and prior research than by the principal components analysis.
10
I took this approach because ICPSR would not release the census data that was collected due to
pending confidentiality issues. They also could not specify when the census data would be available nor
would they provide census track numbers that would allow me to gather the data independently. Hence, to
avoid entirely omitting neighborhood level demographic controls from the models, I decided to aggregate
the level 1 data at the Neighborhood Cluster level. While this approach is not ideal, it is acceptable for
several reasons. First, given the large sample size, NC’s contained an average of 33 (and a maximum of
62) persons each. Second, the Neighborhood Clusters were chosen based on internal homogeneity so
having information about a sizeable handful of people within a cluster should be a rough proxy for
information about the NC more broadly.
123
788
Polit Behav (2013) 35:777–806
variables were modeled at level one and the neighborhood variables were modeled
at level two. Since both of the political participation variables were binary, I
estimated Bernoulli choice random intercept models with a logit link function (Bryk
and Raudenbush 2002: 291–309). I held the slopes constant.11 Specifically, the level
one model was:
Participationij ¼ b0j þ
9
X
q¼1
bqj Xqij þ cij
In this equation, participationij is one of two measures of political participation in
neighborhood j as reported by respondent i; b0j is a neighborhood specific intercept;
Xqij is the value of one of the specified predictors12 (q) associated with respondent i
in neighborhood j; bqj is the partial effect of predictor q on participation and rij is the
individual level random error. The nine predictors (q) at level one include the seven
controls described earlier as well two variables for perceptions: one for the perception scale described above and another for a quadratic term (the square of the
perception scale), which was added to allow for the possibility of non-linear effects.
At level two, the level 1 intercept (b0j) is modeled as follows:
b0j ¼ c00 þ
5
X
q¼1
cqj Xqij þ uoj
In this equation, !00 is the mean participation in a given neighborhood, Xqj is the
value of the predictor q associated with neighborhood j, !qj is the partial effect of
predictor q on mean neighborhood participation and uoj is a neighborhood level
random error. The five level 2 predictors (q) include the SSO scales of objective
physical and social disorder,13 percent black in neighborhood, mean neighborhood
income and mean neighborhood tenure length.14
Findings
The results are presented in Tables 2 and 3. Bearing in mind the number of models,
I will concentrate on the main findings. Each of the tables was organized to illustrate
11
Initially, I considered allowing for the possibility that slope of perceptions might vary based on the
race or class composition of the neighborhood. This would mean that the effect of perceptions would be
different for different kinds of neighborhoods. I tried several models to this end -but there was little
evidence of significant variation in the slopes across neighborhood clusters. Since varying slopes were not
critical to the hypotheses I proposed, I ultimately opted for the simpler and more parsimonious varying
intercept model.
12
In these and all the multilevel models that follow, Level 1 predictors are centered at the group means
to facilitate sound and more easily interpretable empirical results (Bryk and Raudenbush 2002; Enders
and Tofighi 2007).
13
I did not include a quadratic term in the level two model. There is little precedent in the literature for
doing so. However, I nonetheless tested for the possibility that there is a non-linear relationship between
objective disorder and participation and found that this was not the case.
14
In these and all the multilevel models that follow, Level 2 predictors are centered at their grand (or
overall) means (Bryk and Raudenbush 2002; Enders and Tofighi 2007).
123
Polit Behav (2013) 35:777–806
789
Table 2 Estimates predicting speaking to a politician
Model 1 (individual
and neighborhood
demographics)
Model 2
(?objective
observed disorder)
Model 3
(?perceptions
of disorder)
Individual level
Age
.01 (.00)***
.01 (.01)***
.02 (.01)***
Education
.12 (.01)***
.09 (.03)***
.09 (.03)***
Income
.07 (.01)***
.03 (.03)
.04 (.03)
-.10 (.23)
-.28 (.25)
Latino
-.09 (.10)
Black
-.04 (.09)
Mobility
-.15 (.03)***
Home owner
.91 (.07)***
.28 (.27)
.26 (.23)
-.11 (.06)*
-.10 (.06)*
.98 (.18)***
.91 (.20)***
Perceived disorder
–
–
.06 (.03)**
Perceive disorder (quadratic)
–
–
-.02 (.01)***
Neighborhood level
Objective social disorder
–
-.54 (.22)**
-.52 (.24)**
Objective physical disorder
–
-.04 (.04)
-.05 (.04)
Mean income
.18 (.02)***
Proportion black
.01 (.17)
Mean tenure length
.06 (.01)***
.12 (.07)*
-.70 (.47)
.07 (.03)***
.08 (.07)
-.47 (.49)
.06 (.03)**
Level 1 N
7,480
1,191
996
Level 2 N
343
75
74
All models based on multilevel logistic regression
*** p B .01
** p B .05
* p B .10
the results of successive alternative models. Model 1 includes only basic individual
and neighborhood demographics.15 Model 2 adds the scales for objective physical
and social disorder. Finally, Model 3 adds perceptions of disorder along with its
quadratic transformation.16
15
In Tables 2 and 3, the results did not change substantively when I switched the ordering of models 1
and 2. In other words, models with just the basic individual level predictors and the objective measures of
disorder (without demographic neighborhood predictors OR perceptions) showed similar patterns.
16
Note that the sample size decreases dramatically in the move from Model 1 to Model 2 and then again
(although less so) in the move from Model 2 to Model 3. The initial decrease is a result of missing
objective disorder data. Fortunately, the missingness of this data is random, as PHDCN Systematic Social
Observation data was collected and coded only for a random subset of block faces in Chicago and thus
only for a random subset of the people in the survey sample. It is instructive to note that subsetting the
initial models so that they only include observations with full SSO data (e.g. dropping all observations
without full data on SSO variables) produces substantively similar results as those shown in Table 2. The
decrease from Model 2 to Model 3 is more problematic because it is mostly a result of missingness in
response to the perceptions variable, which may not be random. To address this I took two steps. First, I
re-ran the models based on subsetted data that only includes observations with no missingness on key
variables, none of the substantive effects change. Hence, differences in the composition of the sample
123
790
Polit Behav (2013) 35:777–806
Table 3 Estimates predicting attending a meeting
Model 1 (individual
and neighborhood
demographics)
Model 2
(?objective
observed disorder)
Model 3
(?perceptions
of disorder)
Individual level
Age
.01 (.00)***
.00 (.00)
.01 (.01)
Education
.09 (.01)***
.05 (.03)*
.02 (.03)
Income
.05 (.01)***
.05 (.03)*
Latino
Black
Mobility
Home owner
-.06 (.10)
.03 (.09)
-.26 (.23)
.05 (.03)*
-.37 (.25)
-.01 (.21)
-.13 (.23)
-.21 (.03)***
-.18 (.06)***
-.18 (.07)***
.95 (.07)***
1.2 (.19)***
1.3 (.20)***
Perceived disorder
–
–
Perceive disorder (quadratic)
–
–
.06 (.03)**
-.00 (.01)
Neighborhood level
Objective social disorder
–
.16 (.18)
.14 (.19)
Objective physical disorder
–
-.04 (.04)
-.05 (.04)
Mean income
Proportion black
Mean tenure length
.17 (.02)***
-.42 (.17)**
.03 (.01)***
.19 (.06)***
.19 (.07)***
.50 (.45)
.69 (.49)
-.02 (.02)
-.11 (.08)
Level 1 N
7,462
1,184
990
Level 2 N
343
75
74
All models based on multilevel logistic regression
*** p B .01
** p B .05
* p B .10
Results
The empirical results support H1 across the board: perceptions of disorder have
strong positive associations with both forms of local political participation. Since
the perceptions scale is coded such that higher scores represent increasing
perceptions of disorder, this means that as perceptions of disorder increase, the
likelihood of speaking to a politician or attending a meeting also increases.
However, the story does not end there. For speaking to a politician, the quadratic
term (which denotes curvature) is negative and significant. This means that
increasingly negative perceptions are associated with an increasing likelihood of
speaking to a politician at a decreasing rate. Thus, the observed increase is
dampened by the quadratic element of the relationship exerting a downward force
Footnote 16 continued
across models do not appear to drive the results. Second, I imputed the missing observations in the
perceptions scale (based on research referenced above indicating how best to predict perceptions) and ran
the models with this imputed data. Again, none of the substantive effects change. Overall, the sample size
changes and missing data do not appear to bias any of the findings offered in the paper.
123
Polit Behav (2013) 35:777–806
791
Fig. 2 This figure shows a significant linear decrease in the predicted probability of speaking to a
politician as levels of objective social disorder increase (ceteris paribus)
that causes participation to level off and descend. Substantively, this indicates that
neighborhood residents on the extremes—those with a very negative or very
positive perspective towards their environment—are less likely to participate via
formal contact with an elected official than those who fall in the middle. The results
are more mixed with respect to H2. For one of the dependent variables (attending a
meeting), H2 is inaccurate: objective disorder (physical or social) has no direct
effect on participation. However, for the other dependent variable (speaking to a
politician), H2 is partly correct: objective social disorder is a consistently strong
determinant. More precisely, as objective social disorder increases, a respondents’
likelihood of speaking to a politician decreases. Before discussing these findings, it
is helpful to look at the magnitude of the estimated correlations.
Magnitude of Effects
Both subjective perceptions and socially rooted objective markers of disorder have a
substantively sizable association with participation. As shown in Fig. 2, ceteris
paribus, people living in areas with the highest measured levels of objective social
disorder are 23 percentage points less likely to speak to a politician than their
counterparts in neighborhoods with the lowest levels of such disorder. Although
strong, this is the only evidence of a relationship between participation and
objective disorder: social disorder has no discernable impact on the other dependent
variable (attending a meeting) and physical disorder is not a significant predictor of
either outcome.
Contrastingly, subjective perceptions of disorder have strong associations with
both of the dependent variables in question. Respondents with mean perceptions of
disorder are 22 percentage points more likely to speak to a politician than those on
the minimum end of the perceptions range and 15 percentage points more likely
123
792
Polit Behav (2013) 35:777–806
Fig. 3 This figure shows the significant curvilinear relationship between speaking to a politician and
perceptions of neighborhood disorder (note that the scale of the perceptions variable is different than
indicated by the summary statistics due to its subsequent group mean centering)
compared to those on the maximum end (see Fig. 3). These patterns highlight the
curvilinear relationship between perceptions and speaking to a politician.
The story is somewhat different when it comes to meeting attendance. For this
outcome, there is an overall decrease: people who report perceiving the least
disorder (see Fig. 4) are 14 percentage points less likely to attend a meeting than
those who report perceiving the most disorder. There is no significant curvilinear
component of this relationship: respondents with mean perceptions of disorder are
12 percentage points more likely to attend a meeting than those with minimum
perceptions and two percentage points (a statistically indistinguishable difference)
more likely to do so than respondents who perceive the most disorder.
Overall, these findings highlight the extent to which perception is a critical determinant
of individuals’ decisions about whether or not to participate at the local level. Most
notably, perceptions influence participation both linearly (best perceptions = most
participation) and curvilinearly (moderate perceptions = most participation).
Discussion
There are at least three patterns in the findings that compel more detailed discussion:
(1) that a curvilinear relationship exists between speaking to a politician and
perceptions of disorder while a linear relationship exist between attending a meeting
and such perceptions, (2) that objective social disorder has a direct effect on speaking
to a politician yet has no direct effect on attending a meeting, (3) that the coefficients
on objective disorder generally do not change after incorporating perceptions into the
models. Patterns 1 and 2 are closely related insofar as they both have to do with why
we might see divergent results across the two outcomes examined. Although the data
123
Polit Behav (2013) 35:777–806
793
Fig. 4 This figure shows the significant linear increase in probability of attending a meeting as
perceptions of neighborhood disorder increase
does not enable me to explain this with ideal empirical precision, there are three
insights relevant to making sense of it. First, the very motivation for considering
these dependent variables separately was because speaking to a politician and
attending a meeting are fundamentally different approaches to solving local
problems, which we might therefore expect to have distinct determinants. To grasp
the implications of this, it is instructive to examine some of the dissimilarities
between PHDCN respondents who participated by speaking to a politician (only) and
those who participated by attending a meeting (only). Namely, a series of t tests (not
shown) demonstrate that meeting attenders were more socially and organizationally
embedded: they were significantly more likely to be a part of a block group, to have
joined a neighborhood watch, to report having more neighborhood ties and to agree
that most of their neighbors are willing to help one another.
This provides at least a prima facie explanation of why we might see divergent
results between these two outcomes. Those who attend meetings are likely shielded
from the negative effects of social disorder by their connection to more positive
local influences and activities. In the same way, they may be less inclined to turn to
apathy at the extremes of neighborhood perceptions because they are involved in
larger organizational networks that keep them engaged.
Additionally, one reason we might see a special relationship between social
disorder and speaking to a politician (and see no such relationships between social
disorder and attending a meeting or between physical disorder and either of the
dependent variables) is that people ? formal political authority make a more complex
calculus for participation. Engagement with political officials is a distant and less
accessible activity for most people. The PHDCN data confirm this: among the entire
sample, less than three percent of respondents reported belonging to an explicitly
political group (compared to 11 percent who reported belonging to a neighborhood
watch and 12 percent who belonged to a block group). Insofar as inertia works especially
123
794
Polit Behav (2013) 35:777–806
against the act of reaching out to a political representative, it stands to reason that this
form of participation is more vulnerable to the influences of social disorder and more
susceptible to being swayed at the extremes of perceptions of disorder.
Moreover, the residents of blighted neighborhoods might (with good reason) not
feasibly believe that local political agents can do much about drug dealers, drunkards
and the other societal ‘‘deviants’’ that contribute to social disorder. A personal
example, while anecdotal, is illustrative of this argument. I once lived in a distressed
neighborhood on the South side of Chicago. After moving in, I soon discovered that
some of my neighbors were drug dealers. Upon talking to one of the residents of my
building about the issue, I learned that she had inquired with an older woman who
lived nearby about whether someone should approach the alderman of our political
ward to talk about the local drug problem. In response to this inquiry, the older
woman had insisted that such a tactic would not help. Even if the offending parties
were somehow removed, other (perhaps worse) dealers would emerge to meet the
seemingly voracious neighborhood demand for narcotics. Furthermore, making too
much noise might cause officials to take drastic actions, like performing drug raids on
nearby houses and pushing for stiffer sentences for drug dealers. Some of dealers
were nice people with strong local social networks and although the older woman
would prefer that they change their ways, she doubted jail would help. On top of all
this, she noted that one of the local drug runners once worked a good job in
construction and only turned to drugs after hurting his leg and being left with few
other options for taking care of his family. As this (true) story reflects, people (such
as those that lie at the core of social disorder), complicate the participatory equation,
particularly when it comes to deciding whether to mobilize (often distant) political
authorities against (sometimes close) neighbors and friends. Attending a meeting
does not implicate social disorder through political authorities and thus would not tap
into this kind of dynamic. Nor does physical disorder involve people to an extent that
would introduce such challenges.
Abstracting from the details of the example for a moment, the point is this: it
appears that contacting a political official and social disorder are each unique in ways
that explain both why they relate to one another and why the former has a different (i.e.
curvilinear) relationship to perceptions of disorder than does attending a meeting.
As mentioned at the beginning of this section, the third pattern necessitating
explanation was that the coefficients on objective disorder do not change much after
adding perceptions, indicating that the two variables are operating through different
channels in terms of shaping local political participation. This too is reasonable in
light of several things. First, when Sampson and Raudenbush (2004) examine these
relationships in multivariate models they find that after controlling for racial
context, there is no association between perceptions and physical disorder and only
a moderate relationship between perceptions and social disorder. In short, the link
between objective and subjective disorder exists, but it is not overwhelmingly
strong; these measures are each gauging something distinct. This explains how
objective social disorder can effect participation separately from perceptions.
Where and how the mechanisms involved diverge is a tougher question. Given
that objective social disorder only negatively impacts the likelihood of speaking to a
politician (and not that of attending a meeting), I would postulate that it bears on
123
Polit Behav (2013) 35:777–806
795
participation by shaping things like individuals’ relationship to local authority
structures, their sense of the limits of formal politics etc. This would explain how
social disorder could dampen the inclination to speak to a politician but not
necessarily the proclivity for engaging in more grassroots activities like attending a
meeting. Contrastingly, perceptions likely operate on a broader level by guiding the
overarching calculus about the costs and potential benefits of participation writ
large. This elucidates why adding perceptions to the models did not alter the role of
objective disorder. Thinking back to the example I gave earlier is again helpful:
regardless of where my neighbor stood in terms of her perceptions of local disorder,
her decision to advise fellow residents against contacting an alderman was
apparently driven by a distinct mix of other attitudes (i.e. assessments of the
insatiable local demand for drugs, understandings of the larger socioeconomic
forces motivating drug dealers etc.). Some of these factors could be closely
correlated to objective disorder (e.g. being physically proximate to drug dealers and
consumers can lead one to believe that the demand for these things is high and that
the sale and consumption of them is an inevitable and hardly amenable to political
change). If this were the case, the channels through which objective disorder
influenced political participation would lie outside of the mechanism of perceptions.
I have insufficient empirical leverage (and not enough space) for examining these
conjectures more closely. Nevertheless, the findings underscore the need for
political scientists to pay closer attention to the role of objective social disorder in
structuring local participation. It also cautions us against discounting the meaningful
distinctions between different kinds of political behavior.
Alternative Considerations and Robustness Checks
When it comes to causally interpreting this research, the usual caveats apply. While
the findings are strongly suggestive, it is impossible to make unequivocal claims
about cause and effect in the context of a cross-sectional observational study.
However, to bolster the case for the robustness of the results, I performed additional
analyses addressing several potential inferential challenges. I offer detailed
discussions of three such challenges: correlated causes, reverse causality and
critical interactions. As it turns out, empirical scrutiny of these possibilities does not
compromise the core findings.
Correlated Causes
What if an unobserved variable causes both participation and perceptions of disorder?
Omitted variable bias can never be entirely eliminated in the (non-experimental)
survey research setting but attentiveness to the kinds of causes that might
simultaneously account for outcomes and predictors is nevertheless warranted.
Given the subject at hand, two variables stand out in this sense: collective efficacy
and social disorganization. These phenomena are central to the sociological literature
on contextual effects (Sampson and Groves 1989; Sampson et al. 1997; Sampson
et al. 2002). Social disorganization generally refers to, ‘‘the inability of a community
123
796
Polit Behav (2013) 35:777–806
structure to realize the common values of its residents and maintain effective social
controls (Kornhauser 1978: 120; Sampson and Groves 1989). Collective efficacy, a
related concept, lies at the positive end of the social disorganization spectrum.
Namely, collective efficacy is at its height when neighbors, ‘‘get along, work through
local organizations to better the community and take steps to informally control
trouble in their neighborhood (Akers and Jensen 2003: 8).’’
Arguably, individuals who believe that their neighborhoods have low levels of
collective efficacy and/or high levels of social disorganization may also perceive
more disorder. At the same time, individuals who discern a lack of efficacy or
organization in their neighborhoods might be propelled to mobilize their neighbors,
thus spurring the local participation that they believe is absent. The same would be
true at the extreme positive end. Individuals who perceive their neighborhoods to be
highly organized and efficacious will have perceive less disorder and will also have
scant motivation for investing their own political energies (both because they may
not see a need for doing so and because they may take advantage of the opportunity
to free ride off the efforts of their organized and efficacious neighbors). In this way,
collective efficacy and social disorganization could ostensibly cause both perceptions of disorder and political participation. If so, then excluding these things from
the models could threaten the validity of initial findings.
To test for this, I incorporate measures of both phenomena into the originally
estimated participation models. Sociologists have well-established variables to
measure these concepts and I do not reinvent the wheel. Two of the most common
measures of social disorganization are: 1) local friendship networks 2) unsupervised
teenagers/peer groups (Sampson and Groves 1989; Veysey and Messner 1999).
Collective efficacy is generally measured by creating a scale with items tapping into
individuals’ beliefs about: (1) neighborhood social cohesion and (2) social control
(Sampson et al. 1997). The PHDCN has variables and/or scales representing each of
these things.
Tables 4 and 5 contain the results. In both tables, Model 1 shows the original
findings, model 2 shows the impact of adding the two social disorganization variables
and Model 3 shows the results with the addition of the collective efficacy scale. For
both speaking to a politician (Table 4) and attending a meeting (Table 5),
disorganization and efficacy have significant associations with participation.
Respondents who view their neighbors as more organized and more efficacious are
more likely to take political action at the local level. This makes great intuitive sense
because individuals’ sense of neighborhood efficacy and organization presumably
shape their ideas about the larger contexts in which local political engagement plays
out and hence influence the calculus about the potential gains of participation.
The critical point with respect to the main findings is that despite the influence of
efficacy and organization, adding them to the models does not eliminate the
significance of perceptions. The same is true for a host of other potentially
confounding variables included in various iterations of these models (including
individual and neighborhood level social ties, institutional ties, perceptions of
neighborhood danger, perceptions of neighborhood violence, neighborhood attachment and social capital).
123
Polit Behav (2013) 35:777–806
797
Table 4 Models including social disorganization and collective efficacy (predicting speaking to a
politician)
Model 1
(initial full
model)
Model 2
(initial ? social
disorganization)
Model 3
(initial ? collective
efficacy)
Individual level
Age
.02 (.01)***
.02 (.01)***
Education
.09 (.03)***
.10 (.03)***
.10 (.03)***
Income
.04 (.03)
.04 (.03)
.04 (.03)
Latino
-.28 (.25)
-.23 (.25)
-.21 (.25)
Black
.26 (.23)
.32 (.23)
.24 (.23)
-.08 (.06)
-.09 (.06)
Mobility
-.10 (.06)*
.02 (.01)***
Home owner
.91 (.20)***
.99 (.20)***
.82 (.20)***
Perceived disorder
.06 (.03)**
.04 (.04)
.11 (.03)***
Perceive disorder (quadratic)
-.02 (.01)***
-.02 (.01)***
-.02 (.01)***
Objective social disorder
-.52 (.24)**
-.51 (.24)**
-.47 (.24)**
Objective physical disorder
-.05 (.04)
-.06 (.04)
-.06 (.04)
.08 (.07)
.07 (.07)
.05 (.06)
-.47 (.49)
-.44 (.50)
-.35 (.48)
Neighborhood level
Mean income
Proportion black
Mean tenure length
.06 (.03)**
.06 (.03)**
.05 (.03)*
Social disorganization
Friendship networks
–
.24 (.07)***
Unsupervised Teenagers
–
.07 (.06)
–
Collective efficacy scale
–
–
Level 1 N
996
996
996
Level 2 N
74
74
74
–
.50 (.11)***
All models based on multilevel logistic regression
*** p B .01
** p B .05
* p B .10
Reverse Causality
What if local political participation shapes perceptions of neighborhood disorder?
Speaking to a politician or attending a meeting might change one’s perceptions of
the neighborhood environment. If this is the case, then the processes hypothesized
above may not adequately explain the relationship between perceptions of disorder
and participation. There are two reasons that I argue that the causal arrow does not
flow from participation to perceptions. First, the wording of the questions for both
dependent variables creates an inbuilt assumption that the causal ordering goes from
perceptions to participation. Namely, respondents were asked whether they had
spoken to a politician or attended a meeting about a neighborhood problem. This
123
798
Polit Behav (2013) 35:777–806
Table 5 Models including social disorganization and collective efficacy (predicting attending a meeting)
Model 1
(initial
full model)
Model 2
(initial ? social
disorganization)
Model 3
(initial ? collective
efficacy)
Individual level
Age
.01 (.01)
.01 (.01)
Education
.02 (.03)
.03 (.03)
.03 (.03)
Income
.05 (.03)*
.05 (.03)*
.05 (.03)*
Latino
-.37 (.25)
-.40 (.26)
-.40 (.26)
Black
-.13 (.23)
-.07 (.23)
-.17 (.23)
Mobility
-.18 (.07)***
-.15 (.06)**
-.16 (.06)***
Home owner
1.3 (.20)***
1.3 (.21)***
Perceived disorder
.06 (.03)**
.07 (.04)**
Perceive disorder (quadratic)
-.00 (.01)
-.00 (.01)
.01 (.01)
1.2 (.21)***
.26 (.03)***
-.01 (.01)
Neighborhood level
Objective social disorder
.14 (.19)
Objective physical disorder
-.05 (.04)
.16 (.20)
Mean income
.19 (.07)***
.19 (.07)**
Proportion black
.69 (.49)
.73 (.50)
.73 (.50)
Mean tenure length
-.11 (.08)
-.03 (.03)
-.04 (.03)
-.07 (.04)*
.24 (.20)
-.07 (.04)
.16 (.07)**
Social disorganization
Friendship networks
–
Unsupervised teenagers
–
.36 (.07)***
-.00 (.05)
–
–
–
Collective efficacy scale
–
Level 1 N
990
990
.62 (.12)***
990
Level 2 N
74
74
74
All models based on multilevel logistic regression
***p B .01
** p B .05
* p B .10
means that the initial motivation of the participation was constrained to being
related to neighborhood perceptions and thus, that perceptions were causally prior.
Second, while evidence exists linking local participation to perceptions of
neighborhood satisfaction, there is no research indicating a relationship between local
participation and perceptions of disorder (Taylor 2001). The distinction between these
two (perceptions of satisfaction versus perceptions of disorder) is of the utmost
importance here. Since perceptions of satisfaction tap into diverse aspects of the
neighborhood experience, they could ostensibly be shaped by virtually any neighborhood related phenomenon (including local participation). Contrastingly, perceptions of
disorder are more limited, especially as they are measured in the PHDCN data. It is
easier to understand why an encounter with one’s neighbors at a community meeting
could shape broad feelings of residential satisfaction than to understand why it would
123
Polit Behav (2013) 35:777–806
799
more specifically alter one’s ideas about whether litter, vacant housing, drinking in
public or selling drugs are big problems in the neighborhood.
Taking this logic a bit further, it is instructive to envisage the mechanisms
that might form a link from participation to perceptions of disorder, as doing so
gives us empirical leverage for assessing the feasibility of reverse causation.17
One potential mechanism of participation involves the evaluations of one’s
neighbors or neighborhood that develops as a result of either attending a meeting
or speaking to a politician. If participation in these activities were shaping
perceptions of disorder, then prevailing research indicates that they would be
doing so by altering things such as: trust in neighbors, perceptions of whether
neighbors have different values, perceptions of neighborhood close knittedness
and perceptions of whether local authorities were responsive to community
needs. For example, Putnam (2000) argued that participation in local community
organizations creates social cohesion among neighbors and Dassopoulos and
Monnat (2011) found that such cohesion mediates the impact of participation on
perceptions of neighborhood satisfaction. Following such findings, we might
posit that the risk for reverse causation lies in the propensity for local
participation to influence factors related to how people evaluate their neighbors
and neighborhoods (and thus to also affect perceptions of disorder). It further
follows that if endogeneity is driving the previously presented results, then
controlling for the channels through which it might operate should eliminate the
relationship between perceptions of disorder and participation. As the models
presented in Table 6 show, no such thing happens. While circumstantial, this is
modest evidence against reverse causation: even after accounting for a wide
spectrum of the mechanisms through which participation could bear upon
perceptions of disorder-the relationship between the two holds fast.
Critical Interactions
What if the relationship between political participation and neighborhood perceptions varies systematically by class (or some other factor)? That is, what if people
who are wealthier have a different political response to perceptions of disorder than
those who are less privileged? Under such circumstances, the initial findings offered
in this paper could at least partially reflect unaccounted for interactions between
income and perceptions and thus misestimate the direct effect of perceptions.
In a preliminary test for this, I perform basic t tests to determine whether mean
perceptions of disorder differ significantly between low (less than 25 K), middle
(25–60 K) and high (over 60 K) income groups. As it turns out, they do.
17
As a more direct test than this, I initially estimated Two Staged Probit Least Squares (2SPLS) models.
2SPLS is a procedure for estimating non-recursive models when one of the endogenous variables of
interest is dichotomous (Keshk 2003; Alvarez and Glasgow 2000). The results were partially in favor of
the core findings of the article: perceptions of disorder remained significant when the outcome was
attending a meeting, even in a recursive setting. However, this did not hold for the other dependent
variable (speaking to a politician). More generally, the 2SPLS results were questionable (in particular,
they were highly sensitive to model specification). This is likely exacerbated by the multilevel and nonlinear nature of the analysis, which the 2SPLS models I ran were not designed to account for. Given the
uncertainty of these findings, I refrain from using them as evidence for or against reverse causality.
123
800
Polit Behav (2013) 35:777–806
Table 6 Models including potential mechanisms for reverse causation
Speaking to politician
Attending meeting
Individual level
Age
.02 (.01)**
.01 (.01)
Education
.07 (.04)**
.03 (.04)
Income
.04 (.03)
.05 (.03)
-.32 (.28)
-.40 (.28)
Latino
Black
Mobility
.35 (.25)
-.11 (.07)*
Home owner
.81 (.22)***
Perceived disorder
.08 (.03)**
Perceive disorder (quadratic)
-.02 (.01)***
-.11 (.25)
-.19 (.07)***
1.2 (.23)***
.12 (.03)***
-.01 (.01)
Potential mechanisms for reverse causation
Close knit neighborhood
.20 (.09)**
.12 (.09)
Trust neighbors
.16 (.11)
.26 (.12)***
Neighbors have different values
-.09 (.09)
.07 (.09)
Police responsive to local issues
.09 (.10)
.06 (.10)
Neighborhood level
Objective social disorder
-.59 (.26)**
Objective physical disorder
-.06 (.04)
Mean income
Proportion black
.04 (.07)
-.77 (.53)
.16 (.21)
-.06 (.05)
.16 (.08)**
.51 (.55)
Mean tenure length
.06 (.03)**
Level 1 N
784
-.04 (.03)
781
Level 2 N
73
73
All models based on multilevel logistic regression
*** p B .01
** p B .05
* p B .10
To determine whether adjusting for these patterns changes anything, I re-reran
the analyses for each outcome, estimating three separate models, each of which
included interactions between perceptions and one of three income dummy
variables (low, middle or high). I draw on these results to tell me whether the
perceptions-participation slope varies based on the particular income group a
respondent belongs to.
The results (not shown) reveal no such change. This means that whether one is
low-income, middle-income or high-income, the relationship between perceptions
and participation remains significant: curvilinear for speaking to a politician, linear
for attending a meeting.18
18
In addition, the results dis not change when I ran models that simply added a basic interaction between
income and perceptions (as opposed to splitting the interactions up based on income category and running
separate models).
123
Polit Behav (2013) 35:777–806
801
I also checked for interactions between perceptions and education, perceptions
and race and perceptions and objective disorder. None of these were significant and
none of them weakened the relationship between perceptions and participation.
Omitted interaction terms do not appear to be distorting the findings.
Broader Implications
The results detailed throughout this article paint a complex but illuminating picture
of the role of both objective contextual disorder and subjective perceptions of
disorder in shaping individuals’ local political engagement. Laying the groundwork
in partial support of (H2), it turns out that certain kinds of objective disorder seem to
bear on certain kinds of political action: increased levels of objective social disorder
(i.e. that which directly reflects the behavior of people) are associated with a
decreased likelihood of going though formal political channels (speaking to a
politician) in an attempt to solve neighborhood problems. So, while actual broken
windows don’t have much direct political significance, the socially based inclivities
that are part and parcel of Wilson and Kellings’ theory of broken windows do figure
into political decision-making processes.
Even more critical and much in line with my expectations for the other
hypothesis under examination (H1): perceptions of disorder are directly, significantly and strongly associated with political participation. Neighborhood residents
who sense more disorder are more likely to be politically responsive. Moreover, for
speaking to a politician, this association is non-linear; the trend levels off and
descends for people with extreme perceptions. So, neighborhood residents who are
either most or least aware of disorder are decreasingly likely to take this kind of
formal political action. A neighborhood is most conducive to political engagement
when its residents’ perceptions of disorder are severe enough to warrant political
response, although in some cases too much severity may induce hopelessness.
This article furthers both scholarly and practical ends by advancing our
understanding of the relationship between neighborhood disorder (both objective
and subjective) and local political participation. In terms of the former, researchers
have too often attempted to gauge the political implications of neighborhoods
without accounting for disorder: either the concrete sights or sounds of the
neighborhood experience (objective) or the particular ways people perceive them
(subjective). This oversight fuels a larger disconnect between mainstream political
science and the realities of lived experience (Piven and Cloward 1977; Scott 1990;
Peterson 1990; Hardy-Fanta 1993). By focusing on a crucial local institution (the
neighborhood) and basing the analysis of that institution on the, ‘‘substance of
people’s daily lives,’’ I offer a nuanced and textured approach to modeling
contextual effects on political behavior (Verba et al. 1995: 279).
In terms of practical upshots, this work bespeaks the need to remain attentive to
the democratic implications of neighborhood improvement and other local policy
efforts. It is a widely accepted tenet that local political participation can produce
many benefits including improving the quality of the physical environment,
preventing crime and promoting more effective urban service delivery (Rich 1979;
123
802
Polit Behav (2013) 35:777–806
Curtis 1987; Churchman 1987; Florin 1989; Marschall 2004). Given the
consequences of citizen involvement, it is vital that we delineate the processes
that engender it. Yet, amid contentious debates about how to best solve the problems
associated with ‘‘broken windows,’’ there is little (if any) discussion of the
participatory ramifications of disorder. This work brings empirical findings to bear
on these matters. With respect the direct effect of actual broken windows (e.g.
objective physical disorder) on participation, I offer evidence that it is negligible.
However, objective disorder does matter in terms of its social manifestations, so
keeping this kind of disorder at the forefront of the policy agenda is likely to be
better for the political strength of a neighborhood than simply trying to ‘‘clean up’’
blight in the physical sense.
In addition, given the differences in the determinants of attending meetings and
speaking to a politician and the more daunting challenges associated with fostering
the latter, community institutions seeking to bolster participation might benefit from
working to widen the scope of local engagement by promoting access to political
representatives through the same social and organizational networks that sustain
attendance to local meetings. The same is true for local political representatives,
who can cultivate greater community participation by connecting those who reach
out to them to wider organizational circles.
Having said all that, the more compelling value of this research may lie in
demonstrating that perceptions of disorder directly influence local participation.
Turning to perceptions when evaluating policy responses to neighborhood disorder
has potentially far reaching reprecussions. It means we must consider the
participatory implications of things like more aggressive policing strategies and
gentrification policies—deliberating not only about the tangible results of such
policies but also (given what we know about the determinants of perceptions of
disorder) about the ways they may shape residents’ interpretations of disorder.
Previous research clarifies that perceptions of disorder are not just a function of
reality but are instead largely socially constructed and strongly connected to social
psychological mechanisms such as implicit racial and class bias (Sampson and
Raudenbush 2004; Franzini et al. 2008). Since perceptions, ‘‘clearly matter for
reasons that extend far beyond the mere presence of broken windows,’’ it is to
perceptions that we must turn to identify pathways for facilitating local political
participation (Sampson and Raudenbush 2004: 337). In this light, policies
addressing pressing social inequalities are of primary importance. For example,
knowing that racial composition matters for how people perceive neighborhood
disorder, policymakers might pay special attention to housing and education
policies meant to address racial segregation, with a particular eye towards the
potential for such patterns to stymie urban political life.
Local participation is a cornerstone of democratic politics. This article shows that
neighborhood disorder matters for such politics. The degree to which people
perceive graffiti, litter and other signs of disorder as neighborhood problems impacts
their local political responsiveness. This information, combined with broader
understandings of what shapes perceptions of disorder, lays the foundation for
structuring policy in ways that facilitate grassroots activism—a vital component of
the American democratic process.
123
Polit Behav (2013) 35:777–806
803
Acknowledgments This research was made possible by funding from the Ford Foundation and the
University of Chicago. Special thanks go to Cathy J. Cohen, Michael C. Dawson, Christopher A. Bail, Jon
C. Rogowski and the helpful participants of the American Politics Workshop at the University of
Chicago. I also owe a tremendous debt to the editors and anonymous reviewers at Political Behavior for
their valuable insights and comments. All remaining errors are my own.
References
Akers, R. L., & Jensen, G. F. (2003). Social Learning Theory and the Explanation of Crime: A Guide for
the New Century. Advances in Criminological Theory (Vol. 11). New Brunswick, NJ: Transaction
Publishers.
Alex-Assensoh, Y. (1998). Neighborhoods, family and political behavior in urban America. New York:
Gerald Publishing.
Alvarez, R. M., & Glasgow, G. (2000). Two-stage estimation of non-recursive choice models. Political
Analysis, 8, 147–165.
Anderson, M. R. (2009). Beyond membership: A sense of community and political behavior. Political
Behavior, 31, 603–627.
Aneshensel, C. S., & Sucoff, C. A. (1996). The neighborhood context of adolescent mental health.
Journal of Health and Social Behavior, 37(4), 293–310.
Bolan, M. (1997). The mobility experience and neighborhood attachment. Demography, 34, 225–237.
Books, J., & Prysby, C. (1988). Studying contextual effects on political behavior: A research inventory
and agenda. American Politics Quarterly, 16, 211–238.
Brook, D. (2006). The cracks in broken windows. The Boston Globe. Available at: http://www.
boston.com/news/globe/ideas/articles/2006/02/19/the_cracks_in_broken_windows/?page=full.
Brunswik, E. (1956). Perception and the representative design of psychological experiments. Berkeley:
University of California Press.
Bryk, A. S., & Raudenbush, S. W. (2002). Hierarchical linear models: Applications and data analysis
methods (2nd ed.). Thousand Oaks, CA: Sage Publications.
Burbank, M. J. (1995). How do contextual effects work? Developing a theoretical model. In Munroe Eagles
(Ed.), Spatial and contextual models in political research (pp. 166–178). London: Taylor and Francis.
Casciano, R., & Massey, D. (2011). Neighborhood disorder and economic self-sufficiency. Working
paper. Office of Population Research Princeton University.
Chavis, D. M., & Wandesman, A. (1990). Sense of community in the urban environment: A catalyst for
participation and community development. American Journal of Community Psychology, 18, 55–81.
Christie-Mizell, A., & Erickson, R. J. (2007). Mothers and mastery: The consequences of perceived
neighborhood disorder. Social Psychology Quarterly, 70(4), 340–365.
Churchman, A. (1987). Can resident participation in neighborhood rehabilitation programs succeed?
Israel’s project renewal through a comparative perspective. In I. Altman & A. Wandersman (Eds.),
Neighborhood and community environments (pp. 113–162). New York: Plenum Press.
Cohen, C. J., & Dawson, M. C. (1993). Neighborhood poverty and African-American politics. American
Political Science Review, 87, 286–302.
Curtis, L. A. (1987). Policies to prevent crime: Neighborhood, family and employment strategies. Annals
of the American Academy of Political and Social Science, 494, 9–168.
Dassopoulos, A., & Monnat, S. M. (2011). Do perceptions of social cohesion, social support, and social
control mediate the effects of local community participation on neighborhood satisfaction?
Environment and Behavior, 43(4), 546–565.
Eagles, M. (1995). Class versus community? Social identities and political mobilization. In M. Eagles
(Ed.), Spatial and Contextual Models in Political Research (pp. 211–233). London: Taylor and
Francis.
Earls, F. J., Brooks-Gunn, J., Raudenbush, S. W., & Sampson, R. J. (1997). Project on human
development in Chicago neighborhoods: Community survey, 1994–1995. ICPSR02766-v3. Boston,
MA: Harvard Medical School [producer]. Ann Arbor, MI: Inter-university Consortium for Political
and Social Research.
Earls, F. J., Raudenbush, S. W., Reiss, A. J. Jr., & Sampson, R. J. (2002). Project on human development
in Chicago neighborhoods (PHDCN): Systematic social observation, 1995. ICPSR13578-v1.
123
804
Polit Behav (2013) 35:777–806
Boston, MA: Harvard Medical School [producer]. Ann Arbor, MI: Inter-university Consortium for
Political and Social Research [distributor].
Elo, I. T., Mykyta, L., Margolis, R., & Culhane, J. F. (2009). Perceptions of neighborhood disorder: The
role of individual and neighborhood characteristics. Social Science Quarterly, 90, 1298–1320.
Enders, C. K., & Tofighi, D. (2007). Centering predictor variables in cross-sectional multilevel models: A
New look at an old issue. Psychological Methods, 12(2), 121–138.
Epstein, P. D., Coates, P. M., Wray, L. D., & Swain, D. (2006). Results that matter: Improving
communities by engaging citizens, measuring performance and getting things done. San Francisco,
CA: Josey-Bass Publishers.
Florin, P. (1989). Nurturing the grassroots: Neighborhood volunteer organizations and America’s cities.
New York: Citizens Committee for New York City.
Franzini, L., O’Brien, Caughy M, Murray Nettles, S., & O’Campo, P. (2008). Perceptions of disorder:
Contributions of neighborhood characteristics to subjective perceptions of disorder. Journal of
Environmental Psychology, 28, 83–93.
Gay, C. (2012). Moving to opportunity: The political effects of a housing mobility experiment. Urban
Affairs Review, 48(2), 147–175.
Hadley-Ives, E., Stiffman, A. R., Elze, D., Johnson, S. D., & Dore, P. (2000). Measuring neighborhood
and school environments perceptual and aggregate approaches. Journal of Human Behavior in the
Social Environment, 3(1), 1–28.
Haney, T. J. (2007). ‘‘Broken windows’’ and self-esteem: Subjective understandings of neighborhood
poverty and disorder. Social Science Research, 36, 968–994.
Harcourt, B. E. (2001). Illusion of order: The false promise of broken windows policing. Cambridge,
Mass.: Harvard University Press.
Harcourt, B. E., & Ludwig, J. (2006). Broken windows: New evidence from New York City and a fivecity social experiment. University of Chicago Law Review, 73, 271–320.
Hardy-Fanta, C. (1993). Latina politics, Latino politics: Gender, culture and political participation in
Boston. Philadelphia: Temple University Press.
Hauser, R. M. (1974). Contextual analysis revisited. Sociological Methods and Research, 2, 365–375.
Hill, T. D., Ross, C. E., & Angel, R. J. (2005). Neighborhood disorder, psychophysiological distress, and
health. Journal of Health and Social Behavior, 46, 170–186.
Huckfeldt, R. (1986). Politics in context: Assimilation and conflict in urban neighborhoods. New York:
Agathon Press.
Huckfeldt, R., & Sprague, J. (1993). Citizens, contexts, and politics. In A. W. Finifter (Ed.), Political
science: The state of the discipline II (pp. 281–303). Washington, DC: American Political Science
Association.
Kamphuis, C. B. M., Mackenbach, J. P., Giskes, K., Huisman, M., Brug, J., & Van Lenthe, F. J. (2010).
Why do poor people perceive poor neighbourhoods? The role of objective neighbourhood features
and psychosocial factors. Health and Place, 16(4), 744–754.
Kaplan, S., & Kaplan, R. (1982). Cognition and environment: Functioning in an uncertain world. New
York: Praeger.
Kasl, S. V., & Harburg, E. (1972). Perceptions of the neighborhood and the desire to move out. Journal of
the American Planning Association, 38, 318–324.
Kelling, G., & Coles, C. (1996). Fixing broken windows: Restoring order and reducing crime in our
communities. New York: Free Press.
Keshk, O. M. G. (2003). CDSIMEQ: A Program to implement two staged probit least squares. The Stata
Journal, 3(2), 157–167.
Kornhauser, R. (1978). Social sources of delinquency. Chicago: University of Chicago Press.
Kubrin, C. E. (2008). Making order of disorder: A call for conceptual clarity. Criminology and Public
Policy, 7(2), 203–214.
Kweon, B., Ellis, C. D., Lee, S., & Rogers, G. O. (2006). Large-scale environmental knowledge:
Investigating the relationship between self-reported and objectively measured physical environments. Environment and Behavior, 38(1), 72–91.
Lake, R., & Huckfeldt, R. (1998). Social networks, social capital, and political participation. Political
Psychology, 19, 567–584.
Latkin, C. A., & Curry, A. D. (2003). Prospective study of the impact of neighborhood disorder. Journal
of Health and Social Behavior, 44, 34–44.
Lee, B. A., & Campbell, K. E. (1997). Common ground? Urban neighborhoods as survey respondents see
them. Social Science Quarterly, 78, 922–936.
123
Polit Behav (2013) 35:777–806
805
Levitt, S. D., & Dubner, S. J. (2005). Freakonomics: A rogue economist explores the hidden side of
everything. New York: HarperCollins.
Lott, J. (2000). More guns, less crime. Chicago: University of Chicago Press.
Manzo, L. C., & Perkins, D. D. (2006). Finding common ground: The importance of place attachment to
community participation and planning. Journal of Planning Literature, 20, 335–350.
Marschall, M. J. (2004). Citizen participation and the neighborhood context: A new look at the
coproduction of local public goods. Political Research Quarterly, 57(2), 231–244.
Meersman, S. C. (2005). Objective neighborhood properties and perceptions of neighborhood problems:
Using a Geographic Information System (GIS) in neighborhood effects and ageing research. Ageing
International, 30, 63–87.
Oliver, E. J. (2000). City size and civic involvement in metropolitan America. American Political Science
Review, 94, 361–373.
Peterson, S. A. (1990). Political behavior: Patterns in everyday life. Newbury Park: Sage Publications.
Piven, F. F., & Cloward, R. A. (1977). Poor people’s movements: Why they succeed, how they fail. New
York: Pantheon Books.
Pomerantz, J. R. (2003). Perception: Overview. In L. Nadel (Ed.), Encyclopedia of cognitive science (Vol.
3, pp. 527–537). London: Nature Publishing Group.
Putnam, R. D. (1993). Making democracy work: Civic traditions in modern Italy. Princeton, NJ:
Princeton University Press.
Putnam, R. D. (2000). Bowling alone: The collapse and revival of American community. New York:
Simon & Schuster.
Quillian, L., & Pager, D. (2001). Black neighbors, higher crime? The role of racial stereotypes in
evaluations of neighborhood crime. American Journal of Sociology, 107, 717–767.
Rich, R. C. (1979). The roles of neighborhood organizations in urban service delivery. Journal of Urban
Affairs, 1, 81–93.
Roberts, S. (2011). Fewer broken windows, and a decline in crime. The New York Times. Available Online:
http://cityroom.blogs.nytimes.com/2011/02/23/fewer-broken-windows-and-a-decline-in-crime/.
Rohe, W. M., & Basolo, Victoria. (1997). Long-term effects of homeownership on the self-perceptions
and social interaction of low-income persons. Environment and Behavior, 29, 793–819.
Rohe, W. M., & Stegman, M. A. (1994). The impact of home ownership on the social and political
involvement of low-income people. Urban Affairs Review, 30(1), 152–172.
Rosenstone, S. J., & Hansen, J. M. (1993). Mobilization, participation and democracy in America.
London: Macmillan Press.
Ross, C. E. (2000). Neighborhood disadvantage and adult depression. Journal of Health and Social
Behavior, 41(2), 177–187.
Ross, C. E., Mirowsky, J., & Pribesh, S. (2001). Powerlessness and the amplification of threat:
Neighborhood disadvantage, disorder, and mistrust. American Sociological Review, 66, 568–591.
Sampson, R. J., & Groves, W. B. (1989). Community structures and crime: Testing social-disorganization
theory. American Journal of Sociology, 94, 774–802.
Sampson, R. J., Morenoff, J. D., & Gannon-Rowley, T. (2002). Assessing ‘neighborhood effects’: Social
processes and new direction in research. Annual Review of Sociology, 28, 443–478.
Sampson, R. J., & Raudenbush, S. W. (1999). Systematic social observation of public spaces: A new look
at disorder in urban neighborhoods. American Journal of Sociology, 105, 603–651.
Sampson, R. J., & Raudenbush, S. W. (2004). Seeing disorder: Neighborhood stigma and the social
construction of ‘Broken Windows’. Social Psychology Quarterly, 67, 319–342.
Sampson, R. J., Raudenbush, S. W., & Earls, F. (1997). Neighborhoods and violent crime: A multilevel
study of collective efficacy. Science, 277, 918–924.
Santiago, A. M., Galster, G. C., Kaiser, A. A., Roman, A. H. S.-S., Grace, R. A., & Linn, A. T. W. (2010).
Low-income homeownership: Does it necessarily mean sacrificing neighborhood quality to buy a
home? Journal of Urban Affairs, 32, 171–198.
Scott, J. C. (1990). Domination and the arts of resistance: Hidden transcripts. New Haven: Yale
University Press.
Skogan, W. G. (2008). Broken windows: Why-and How -We Should Take Them Seriously. Criminology
and Public Policy, 7(2), 195–201.
Skogan, W. G. (2012). Disorder and Crime. In Brandon C. Welsh & David P. Farrington (Eds.), The
Oxford handbook of crime prevention (pp. 173–188). New York: Oxford University Press.
Small, M. L. (2004). Villa victoria: The transformation of social capital in a Boston barrio. Chicago:
University of Chicago Press.
123
806
Polit Behav (2013) 35:777–806
Steenbergen, M. R., & Jones, B. S. (2002). Modeling multilevel data structures. American Journal of
Political Science, 46, 218–237.
Taylor, R. B. (2001). Breaking away from broken windows. Boulder, CO: Westview Press.
Verba, S., Schlozman, K. L., & Brady, H. (1995). Voice and equality. Cambridge: Harvard University
Press.
Veysey, B. M., & Messner, S. F. (1999). Further testing of social disorganization theory: An Elaboration
of Sampson and Groves’s ‘‘Community Structure and Crime’’. Journal of Research in Crime and
Delinquency, 36, 156–174.
Wandersman, A., Jakubs, J. F., & Giamartino, G. A. (1981). Participation in block organizations. Journal
of Community Action, 1, 40–48.
Weden, M. M., Carpiano, R. M., & Robert, S. A. (2008). Subjective and objective neighborhood
characteristics and adult health. Social Science and Medicine, 66(6), 1256–1270.
Wilson, J. Q., & Kelling, G. (1982). The police and neighborhood safety: Broken windows. The Atlantic
Monthly, 127, 29–38.
123